The second implication of these studies is that for parents who withdraw their children from the labor force and send them to school fertility is significantly lower. Since work and school for a child are neither mutually exclusive nor necessarily exhaustive of their time, analysis should deal with the allocation of the child's time to both activities. For Puerto Rico and Egypt, both child labor (unpaid family workers) and school attendance, or education, are treated together in the regression analysis, and both have the expected direct and inverse associations with fertility. In Colombia and Taiwan, only child school attendance rates can be measured, and in both countries the enrollment rate is powerfully inversely associated with fertility.*
To interpret this seemingly general association, more information is required on what determines interregional differences in school attendance rates. Is it the limited capacity of the school system, the supply of schooling, that is restricting attendance? Or is it the willingness of parents to send their children to receive basic education, the demand for schooling, that determines the interregional variation in school attendance rates? The more plausible assumption, as argued earlier, is that although the capacity of school systems, particularly in rural areas, may constrain the acceptance of students in the short run, enrollment rates in the long run reflect the parents' willingness to demand schooling and invest in the education of their offspring.
Adult education may also play a role in facilitating the fall in birth rates, but on this point the findings are mixed. For Colombia and Puerto Rico, adult educational attainment is negatively associated with fertility. But in Taiwan and the Philippines, the association is weak and ambiguous. In Egypt, where a more complex formulation of the model was considered, it appears that women's education contributed to a higher female labor force participation rate, which was in turn associated with somewhat lower fertility.
Neither agricultural employment nor rural residence was in itself a help in accounting for interregional or interpersonal differences in birth rates. Poor health and limited educational opportunities of the agricultural and rural populations fully explain the tendency for these groups to have higher birth rates than those engaged in nonagricultural employment or residing in urban areas. Therefore, interregional and mtersectoral migration per se need not influence reproductive behavior.
Much more research is needed to specify adequately a multivariate model of family reproductive and economic behavior in simple terms and to estimate its parameters for a variety of populations from both aggregate and individual family data. Yet, these initial investigations point to important areas of interaction between the parents' environment and their reproductive and eco-
*Caldwcll (39, 40) reports a thoughtful and thorough discussion of the interrelationships between education and fertility preferences in regions of Ghana. Jones (in this volume) elaborates a variety of further possible interactions between the educational process and fertility.